Databricks Unity Catalog MCP Server
Access your Databricks workspace through Claude and other LLMs. Query Unity Catalog tables, inspect jobs, and retrieve detailed metadata.
README
Databricks Unity Catalog MCP Server
Access your Databricks workspace through Claude and other LLMs. Query Unity Catalog tables, inspect jobs, and retrieve detailed metadata—all through the Model Context Protocol.
Built on the Databricks SDK to provide read-only access to your workspace through the Model Context Protocol. Powered by FastMCP with async/aiohttp for efficient parallel data retrieval.
Read more about our vision and use cases here.
Table of Contents
Features
Capabilities
What you can do:
- Ask Claude to find tables in your Unity Catalog
- Inspect job configurations and recent runs
- Generate queries based on your schema
Limitations
What you can't do:
- Modify tables or jobs (read-only by design)
- Execute queries directly (retrieves metadata only)
Available Tools
Unity Catalog
| Tool | Description | Parameters |
|---|---|---|
get-all-catalogs-schemas-tables |
List all tables across catalogs and schemas | None |
get-table-details |
Retrieve table descriptions, columns, and metadata | full_table_names (list of catalog.schema.table) |
Jobs
| Tool | Description | Parameters |
|---|---|---|
get-jobs |
List all workspace jobs with IDs and names | None |
get-job-details |
Get job settings, configurations, and tasks | job_ids (list of job IDs) |
get-job-runs |
Fetch recent run history with duration, parameters, and results | job_ids (list), n_recent (1-5, default: 1) |
Quick Start
Prerequisites:
- Docker Desktop installed and running
- Databricks workspace access (host URL and access token)
Installation
Choose your editor and follow the configuration steps:
<details> <summary><strong>Cursor</strong></summary>
<br>
Step 1: Add the following configuration to .cursor/mcp.json:
{
"mcpServers": {
"databricks": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"DATABRICKS_HOST",
"-e",
"DATABRICKS_TOKEN",
"ghcr.io/revodatanl/databricks-mcp-server:latest"
],
"env": {
"DATABRICKS_HOST": "${env:DATABRICKS_HOST}",
"DATABRICKS_TOKEN": "${env:DATABRICKS_TOKEN}"
}
}
}
}
Note: You can either use environment variable references (
${env:VARIABLE}) or hardcode the values as strings directly in the configuration.
Step 2: Create a .env file in your project root with your credentials:
DATABRICKS_HOST=your-workspace-url
DATABRICKS_TOKEN=your-access-token
Step 3: Restart Cursor to load the MCP server.
Step 4: Use the cursor rules to enhance your Databricks development workflow.
Learn more about MCP in Cursor
</details>
<details> <summary><strong>Continue.dev</strong></summary>
<br>
Step 1: Add the following configuration to .continue/mcpServers/databricks-mcp.yaml:
name: databricks_mcp_server
version: 0.1.3
schema: v1
mcpServers:
- name: databricks_mcp_server
command: docker
args:
- run
- -i
- --rm
- -e
- DATABRICKS_HOST=${{ inputs.DATABRICKS_HOST }}
- -e
- DATABRICKS_TOKEN=${{ inputs.DATABRICKS_TOKEN }}
- ghcr.io/revodatanl/databricks-mcp-server:latest
Step 2: Set your credentials either:
-
On the Continue.dev website (recommended for security)
-
Or in a
.envfile in your project root:DATABRICKS_HOST=your-workspace-url DATABRICKS_TOKEN=your-access-token
Step 3: Restart your editor to load the MCP server.
Step 4: Use the Continue.dev rules to enhance your Databricks development workflow.
Learn more about MCP in Continue.dev
</details>
Local Development
For contributors and developers who want to run the server locally:
Setup
-
Install uv - Fast Python package installer Follow the installation guide
-
Clone the repository
git clone https://github.com/revodatanl/databricks-mcp-server.git cd databricks-mcp-server -
Install dependencies
uv sync -
Set environment variables
export DATABRICKS_HOST=your-workspace-url export DATABRICKS_TOKEN=your-access-token -
Run the server
uv run databricks-mcp
License
MIT License - see LICENSE.md for details.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。